Heterogeneous graph convolutional network with local influence

Graph convolutional networks (GCNs) have recently drawn extensive attention due to their superior learning performance on graph data. Through graph convolution, topological structure and node attributes can be simultaneously aggregated in a local neighborhood. In heterogeneous information networks (HINs), the diversity of node and edge types poses great challenges to graph convolution. This paper proposes a heterogeneous graph convolutional network based on local influence (named HIGCN), which aims to discriminatively aggregate structural information, attribute information and multi-semantic information in HINs. Here, local influence refers to the influence of neighborhood nodes on the central node. Firstly, a HIGCN block is constructed, in which the local influence is calculated through a heuristic structural influence strategy proposed in this paper and an attention-based attribute influence strategy. Afterwards, 1 × 1 convolution is innovatively used to fuse the embeddings under multiple semantics. Finally, the entire HIGCN framework is constructed by stacking HIGCN blocks. Experiments on real-world network datasets show that HIGCN achieves higher accuracy than related methods in various downstream tasks (node classification, link prediction, etc.), which verifies the effectiveness of the structural influence strategy and the semantic fusion method.